The lignin powders (Sigma-Aldrich) were soaked in 2 M hydrochloric acid solution (mass to volume ratio, 1 g per 20 mL) and stirred for 2 h. After that, sodium hydroxide solution was used to neutralize the above mixture, followed by filtration and freeze drying. The DS of SPEEK was controlled to about 41, 50, 59, and 72%, which were measured by titration method according to our previous work (Jia et al., 2015a). For the preparation of SPEEK/lignin composite membranes, 1.6 g of SPEEK was dissolved in 50 mL dimethyl sulfoxide (DMSO), and stirred at 60°C until being dissolving completely. After cooling down to room temperature, 240 mg pretreatment lignin powder was added and vigorously stirred overnight. The mixture was casted on a home-made glass plate and dried in oven at 100°C for solvent evaporation. After cooling, the membrane was peeled off from the glass and soaked in water immediately. Based on the mass ratio of lignin to SPEEK, the composite membranes were named as SPEEK41/L15, SPEEK50/L15, SPEEK59/L15, and SPEEK72/L15. Commercial Nafion 212 membrane (DuPont) was used as the reference. All of the reagents were used as received.

Fourier transform infrared spectroscopy (FT-IR, Thermo Fisher Nicolet iS10) was investigated in the range of 400–4,000 cm⁻¹. The microstructure of the as-prepared membranes was examined with field-emission scanning electron microscopy (FESEM, Zeiss Auriga FIB/SEM). Water uptake (WU) was measured as follow: (1) dry membrane was soaked into distill water for 24 h; (2) the membrane was taken out and cleaned with filter paper immediately; (3) the cleaned membrane was weight by Mettler-Toledo analytical balance (ME204E). Swelling ratio (SR) was obtained by measuring the length variation of the membrane before and after immersing in deionized water for 24 h. The WU and SR of these membranes can be calculated through the following equations (Ling et al., 2012).

where the W_dry and W_wet are the weight of the membrane before and after soaking, respectively; the L_dry and L_wet are the length of the membrane before and after soaking, respectively.

The permeability of VO²⁺ ion across the membranes was investigated as follow: (1) isolating two reservoirs that were filled 70 mL of 1.5 M VOSO₄ in 3.0 M H₂SO₄ solution and 70 mL of 1.5 M MgSO₄ in 3.0 M H₂SO₄, respectively, with a membrane of 2.01 cm² active area; (2) stirring continuously and measuring the concentration of VO²⁺ in MgSO₄ compartment at 24 h intervals by TU-1900 UV-vis spectrometer; (3) sampling with replacement to keep the solution volume stable. A typical experimental setup was shown in Figure 2. The permeability value of the membrane can be calculated using Equation (3) (Jia et al., 2015a):

where V, A, P, L, C, and C(t) are the volume of the VOSO₄ solution, the effective area of the membrane, the permeability of the vanadium ions, the thickness of the membrane, the initial concentration of VO²⁺ in the VOSO₄ compartment, and the vanadium concentration in the MgSO₄ compartment at the moment t, respectively.

The area resistance (R) of membranes was investigated by a resistance tester (DME-20, DM, China). The electrolytes in both compartments were 1.5 M VOSO₄ and 3 M H₂SO₄, and the conductivity (σ) of the membrane can be calculated as follow (Ling et al., 2012):

where R represents the resistance difference between the cell with and without membrane, L is the thickness of the membrane, and A is the active area of membrane (13.5 cm²).

The VRB single cell consisted of a composite membrane (13.5 cm²) sandwiched between two carbon felt electrodes (13.5 cm²), and two graphite polar plates (current collectors). The 1.5 M VO2+ in 3.0 M H₂SO₄ and 1.5 M V³⁺ in 3.0 M H₂SO₄ solutions were used as catholyte and anolyte, respectively. The cell performance was measured by Arbin battery testing system (BT-I, Arbin, USA) including open circuit voltage decay (OCV), long cycle charge-discharge, and rate performance. The OCV (75% state of charge) was terminated when the voltage of the testing cell declined below 0.85 V. The potential range was between 0.7 and 1.75 V at room temperature.

Figure 3 shows the FT-IR spectra of lignin, pure SPEEK, and the composite membranes. The broad band at 3,400 cm⁻¹ is ascribed to the hydrogen bond and OH vibration. For lignin, the peaks at 2,938 and 2,849 cm⁻¹ are assigned to the C-H stretch in methyl and methylene group (-CH₂-), and those at 1,594 and 1,511 cm⁻¹ are attributed to the aromatic rings of phenyl propane skeleton (characteristic bands of lignin; Faix, 1991). For SPEEK membrane alone, the following fingerprint absorption peaks are present: 3,428 cm⁻¹ (O-H stretching of -SO₃H groups), 1,076 cm⁻¹ (symmetric stretching of O = S = O), and 706 cm⁻¹ (S-O stretching; Li et al., 2014b; Ma et al., 2018). For composite membranes, the intensity of the peaks for -SO3- (1,076, 1,020, and 706 cm⁻¹) increases with the DS, because of the increased amount of the -SO₃H. More importantly, all of the composite membranes exhibit the characteristic peaks of lignin, indicating the incorporation of lignin into the SPEEK matrix. Compared with the pure SPEEK, the peaks of O-H stretching (3,248 cm⁻¹) shift toward the low frequency and the peak intensity decreases in the composite membranes. This suggests the hydrogen bond interaction between the -OH groups of lignin and the -SO₃H groups of SPEEK.

Figures 4A,B reveals that the lignin powders are homogeneously dispersed in the SPEEK solution without any precipitates, even after 240 h. This is a prerequisite for forming a uniform composite membrane. Figures 4C,D displays the resulting SPEEK and SPEEK50/L15 membranes which exhibit uniform and dense surface. This would hamper the crossing of vanadium ions and enhance the cycle stability of the membrane. Moreover, the lignin particles are homogeneously embedded into the polymer matrix and can improve the wetting property of the composite membrane. The 3D structure of lignin also provides more pathways for protons transport and reduces the water channels to repulse vanadium ions. Therefore, the lignin is introduced into the SPEEK matrix, which can enhance the proton conductivity while inhibit the vanadium ions permeation.

The proton transport generally proceeds via the vehicle and Grotthuss mechanism in the membrane. As such, the water uptake is a critical property. It has been demonstrated that the proton conductivity of the proton exchange membrane enhances with the WU amount. However, a high water uptake usually results in low mechanical stability. Table 1 shows that the WU of SPEEK/lignin membranes increases with the DS, as more hydrophilic -SO₃H groups enhance the wetting property of the membrane. Similarly, the SR displays the same tendency as the WU.

The permeability of the resulting membranes is shown in Figure 5. The VO²⁺ permeability of SPEEK/lignin membranes increases with the degree of sulfonation. This is attributed to the high DS that generally imparts abundant -SO3- groups in the polymer matrix, which drastically improve the proton conductivity (see Table 1) and also accelerate the crossover of vanadium ions through the membrane. The ion selectivity (Zhang et al., 2014a; Ji et al., 2017), namely the ratio of proton conductivity to ion permeability, is widely used to describe the balance between the two processes. Figure 5B shows that the ion selectivity of the composite membranes is much higher than that of the Nafion 212, regardless of the DS in SPEEK, because of the suppressed permeability of VO²⁺ ion. This clearly demonstrates the advances of the composite system. In particular, the SPEEK59/L15 membrane exhibits the best performance (61.96 × 10⁴ S min cm⁻³), nearly five-fold increment as compared to the Nafion 212 (12.78 × 10⁴ S min cm⁻³). The incorporation of lignin into the SPEEK substrate provides more proton transport pathway and meanwhile adjusts the water channels to repulse vanadium ions, thus giving rise to high ion selectivity.

To systematically study the performance of the composite membranes, the OCV, long cycle process, and rate performance of VRB single cells were carried out under different conditions. The OCV is a critical parameter to verify the vanadium ions cross rate in the membranes, as the vanadium ions crossing the membrane results in self-discharge and therefore the cell voltage declines accordingly. Figure 6A reveals that the OCV curves decrease rapidly with the enhanced DS in the composite membrane. It is obvious that the voltage decay (above 0.85 V) of cells assembled with the composite membranes is much slower than that of Nafion 212 (288.6 vs. 14.05 h for SPEEK41/L15 vs. Nafion). This indicates that the SPEEK/lignin composite membrane efficiently suppresses the permeation of vanadium ions, in good agreement with the results in Figure 5.

Figure 6B displays the typical charge-discharge curves of the VRB cells using different membranes under current density of 120 mA cm⁻². As a high area resistance (see Table 1) usually leads to high ohmic polarization, the average charge voltage of cells with SPEEK/lignin membranes is slightly higher than that of the Nafion 212, with the average discharge voltage being on the contrary. However, the SPEEK59/L15 membrane exhibits the best discharge capacity, as a result of high ion selectivity, consistent with the above discussion.

Figure 7 shows the cycling performance of cells assembled with SPEEK59/L15 and Nafion 212 membranes from 50 to 320 mA cm⁻². The Coulombic efficiency of the cells with SPEEK59/L15 (up to 99.56%) is higher than that of Nafion 212 over the whole rate range. This originates from the reduced vanadium ion permeability with the composite membranes. It is noted that the CE of both cells increases with the current density. This is mainly due to the shortened charge-discharge time under high current density, which suppresses the crossover of vanadium ions through the membrane. Figure 7B reveals that the discharge capacity of SPEEK59/L15 is higher than that of the Nafion 212 under current density ≤ 250 mA cm⁻². This can be attributed to the overpotential and ohmic polarization under high current. Therefore, the better performance of SPEEK59/L15 is ascribed to the synergistic effect from the lignin additives and SPEEK matrix, which provides more proton transport pathway and meanwhile suppresses the crossover of the vanadium ions.

Figure 8 displays the stability performance of the corresponding VRB cells operated at 120 mA cm⁻². The Coulombic efficiency decreases with the degree of sulfonation (Figure 8A). This agrees well with the varying trend of VO²⁺ permeability (Figure 5A), as a high DS improves the proton conductivity but accelerates the crossover of vanadium ions through the membrane. As such, the voltage efficiency increases with the DS (Figure 8B). Accordingly, the cells with SPEEK59/L15 membranes exhibit an impressive energy efficiency (EE = CE × VE, up to 82.75%), comparable to that of the Nafion 212. This is attributed to the best ion selectivity for the SPEEK59/L15 membrane. It is noteworthy that the cells with SPEEK72/L15 fail after only 29th cycles, as a high DS deteriorates the mechanical and chemical stability of the membrane.

The capacity retention of VRB cells is a key factor to measure the ion imbalance in the operation process. Figure 8D discloses that the discharge capacity retention of the cells assembled with Nafion 212 maintains just 55.96% after 100 cycles under 120 mA cm⁻². In contrast, under the same conditions, the cells with SPEEK59/L15 keep about 94.80% after 100 cycles, and more than 85% after 200 cycles. This demonstrates that the SPEEK59/L15 membrane substantially suppresses the crossover of vanadium ions and thus enhances the stability and prolongs the cycle life. The cells with the SPEEK59/L15 membranes show the best capacity retention and outperform the other ones. The SPEEK59/L15 membranes synthesized by an eco-friendly and cost-effective approach exhibits high ion selectivity and excellent stability, making it a promising candidate for efficient VRB system.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.